Reinforcement learning and decision making via single-photon quantum walks